Potential conflict of interest: Nothing to report.
Five to 15% of patients with primary sclerosing cholangitis (PSC) develop cholangiocarcinoma (CC) with a median survival of 5 to 7 months, an outcome not significantly improved by liver transplantation. However, if CC is found incidentally during the procedure or in the explanted liver, 5-year survival rates of 35% are reported. A noninvasive method to detect CC small enough to allow for intended curative surgery is needed. Unfortunately, computed tomography (CT) and ultrasonography (US) have poor sensitivity for detection of CC in PSC; however, positron emission tomography (PET) using 2-[18F]fluoro-2-deoxy-D-glucose (FDG) differentiates well between CC and nonmalignant tissue. We examined whether PET findings are valid using a blinded study design comparing pretransplantation FDG-PET results with histology of explanted livers. Dynamic FDG-PET was performed in 24 consecutive patients with PSC within 2 weeks after listing for liver transplantation and with no evidence of malignancy on CT, magnetic resonance imaging, or ultrasonography. The PET Center staff was blinded to clinical findings, and surgeons and pathologists were blinded to the PET results. Three patients had CC that was correctly identified by PET. PET was negative in 1 patient with high-grade hilar duct dysplasia. In 20 patients without malignancies, PET was false positive in 1 patient with epitheloid granulomas in the liver. In conclusion, dynamic FDG-PET appears superior to conventional imaging techniques for both detection and exclusion of CC in advanced PSC. FDG-PET may be useful for screening for CC in the pretransplant evaluation of patients with PSC. (HEPATOLOGY 2006;44:1572–1580.)
Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease characterized by inflammation and fibrosis of the intrahepatic and extrahepatic bile ducts. It predisposes to development of cholangiocarcinoma (CC) and possibly even hepatocellular carcinoma. The prevalence of CC in PSC is 5% to 15%.1–5 CC complicating PSC is usually diagnosed at an advanced tumor stage, with an estimated median survival of 5 to 7 months.6, 7 Prognostic models,1, 4, 7, 8 carbohydrate antigen (CA 19-9), carcinoembryonic antigen,9, 10 and concurrent extensive ulcerative colitis or ulcerative colitis with colorectal dysplasia2 are indicators of increased risk of CC in groups of patients with PSC, but such markers are less useful for predicting risk in individual patients.
Survival after liver transplantation of patients with established CC is poor, and most transplantation centers do not accept such patients for liver transplantation. However, 5-year survival rates of 35% following transplantation for PSC with suspected or perioperatively diagnosed CC have recently been reported.11 Unfortunately, diagnostic sensitivity and specificity of conventional imaging techniques—including ultrasonography (US), computerized tomography (CT), and magnetic resonance imaging (MRI)—are not sufficient for diagnosing CC at such an early stage that curative treatment or improved survival after surgical resection or liver transplantation can be achieved. A noninvasive method for detecting small (and early) CC tumors in patients with PSC or exclusion of CC is much needed for both individual pretransplantation evaluations and screening.
US, CT, and MRI yield images with great anatomical accuracy, whereas positron emission tomography (PET) yields images of biochemical processes. These methodological differences and increased glucose metabolism in many tumor cells constitute the background for the generally improved sensitivity and specificity of tumor detection by PET using the glucose analog 2-[18F]fluoro-2-deoxy-D-glucose (FDG) compared with CT and US.12, 13 This holds true for CC,14–17 and in a previous PET pilot study of patients with PSC and established CC and PSC patients without malignancy,18 we found that dynamic FDG-PET consistently differentiated malignant from nonmalignant tissue.
The aim of this study was to prospectively evaluate the usefulness of dynamic FDG-PET for detection of CC not visible on US, CT, or MRI in patients with PSC listed for liver transplantation. Using a blinded study design, PET results and histopathology of the explanted livers were compared in each patient.
The study was a prospective exploratory study of consecutive patients with PSC from liver centers in Sweden and the western half of Denmark. The patients were accepted for liver transplantation from March 2000 to May 2005 according to current criteria in Scandinavia.11 PET examinations were performed at the PET Center of Aarhus University Hospital in Denmark within 2 weeks after listing for transplantation. The only information available for the PET Center was patient identification and that the patient was listed for liver transplantation for PSC. Liver transplantation was performed at Sahlgrenska University Hospital in Gothenburg or Karolinska University Hospital in Huddinge, Stockholm (both in Sweden) or at Copenhagen University Hospital in Denmark. No results of the PET examination were available at transplantation. The explanted livers were scrutinized using serial sectioning supplemented with detailed examinations of biliary strictures. The pathologists were blinded to the results of the PET examination.
Under the assumption of a 10% risk for incidental CC, we aimed to include 30 patients. This would on average yield 3 patients with CC and 27 patients without CC. Sensitivity, specificity, and negative and positive predictive values were estimated with exact confidence intervals based on the binomial distribution.19 The populations of Sweden and of the western part of Denmark were 8.8 million and 2.5 million, respectively, allowing for inclusion of 30 patients within 4 years assuming an unchanged transplantation rate for PSC in Scandinavia.11
Patients and Pretransplantation Workup.
Twenty-four patients were included in the study (20 men and 4 women). Clinical data are shown in Table 1. Indications for transplantation were end-stage cirrhosis due to PSC (14 patients) and advanced biliary strictures not manageable using further dilatation and stenting procedures (10 patients). At the time of transplantation, 6 patients had major hilar strictures, 4 patients had biliary stents, and 3 patients had percutaneous transhepatic drainage of the biliary tree (Table 2).
Table 1. Clinical Characteristics of 24 Patients With PSC Listed for Liver Transplantation
NOTE. Continuous variables are given as the median (range); other values are given as numbers.
Pretransplantation examinations included US and CT in all patients, supplemented with MRI in 8 of 24 (33%) patients. Further visualization of the bile ducts using endoscopic retrograde cholangiography, magnetic resonance cholangiography, and/or percutaneous transhepatic cholangiography was performed in 19 of 24 patients (79%) with biochemical or CT evidence of biliary strictures (Table 2). Brush cytology (endoscopic retrograde cholangiography) in 5 of 11 patients revealed atypical biliary epithelial cells (Table 2). In one of these patients (patient 9), suspected adenocarcinoma was found in biopsies from a hilar stricture, but because subsequent laparoscopy showed no evidence of local invasion or metastatic spread, she was listed for transplantation, which was performed 88 days later. The tumor marker carbohydrate antigen 19-9 was slightly elevated in 10 patients (Table 2). No definite evidence of malignant disease was found in any patient at acceptance for transplantation.
Twenty-three patients underwent transplantation. One patient (patient 11) with no evidence of malignancy at laparoscopy before listing for transplantation was withdrawn 96 days after listing when repeated laparoscopy showed peritoneal carcinosis.
Eleven eligible Swedish patients were not included in the study. Eight of those patients did not wish to travel to Aarhus for PET scanning, 1 patient underwent transplantation 6 days after listing and before he was scheduled for PET examination, 1 patient received intravenous antibiotics for cholangitis, and 1 patient was bleeding from esophageal varices.
The study was approved by the Ethics Committees for each participating center (Aarhus University Hospital and the Universities of Gothenburg, Linköping, Lund, Umeå, Uppsala, Örebro, and the Karolinska Institute in Stockholm). Written informed consent was obtained from each patient.
Scanning was performed at the PET center at Aarhus University Hospital using a Siemens ECAT EXACT HR PET tomograph (CTI/Siemens, Knoxville, TN). FDG was produced at the PET Center applying standard techniques and commercially available systems (General Electric, Uppsala, Sweden).
The patients were scanned in the morning after fasting overnight. Blood glucose levels were 3.9-6.8 mmol/L (mean 5.3 mmol/L). Each patient was positioned in the scanner with the liver within the 15-cm field of view of the scanner. After a 15-minute transmission scan using external sources, an intravenous injection of 300 MBq FDG was given. Dynamic emission recordings were started at the time of tracer injection and comprised 38 time frames: 18 × 10 seconds, 4 × 30 seconds, 5 × 60 seconds, 6 × 300 seconds, and 5 × 600 seconds, for a total of 90 minutes. Data were recorded as mean values in each time frame and were corrected for attenuation based on the transmission scan and radioactive decay to start of the scan. Reconstruction comprised a two-dimensional back-projection algorithm, resulting in three-dimensional images consisting of 128 × 128 × 47 voxels of 2.0 × 2.0 × 3.1 mm3. Central spatial resolution was 6.7 mm (full-width at half-maximum).
A short catheter (Artflon; Becton Dickinson, Swindon, UK) was placed percutaneously in a radial artery under local anesthesia (Lidocain; Danish Hospitals Pharmacy, Copenhagen, Denmark). During the scans, 29 arterial blood samples were collected manually at the following time points after tracer injection: 18 × 5 seconds, 7 × 30 seconds, 1 × 150 seconds, and 8 × 600 seconds. Blood radioactivity concentration was measured using a well counter (Packard Instruments Co., Meriden, CT) that was cross-calibrated with the tomograph and was corrected for radioactive decay to start of the scan. Figure 1 shows an example of the time courses of radioactivity concentrations in tumor tissue, liver tissue, and arterial blood.
No complications to the procedures were seen in any patient. The average radioactivity dose received by the patients from the PET scanning was 5.7 mSv.
Static scan images, constructed as images of the mean tissue radioactivity concentrations 60 to 90 minutes after FDG injection, were examined for focal areas with increased radioactivity concentration compared with surrounding liver tissue.12
Dynamic scan images were constructed as parametric images of the net metabolic clearance of FDG, K (mL blood/min/cm3 tissue), as follows: the relationship of the time course of radioactivity concentrations in liver tissue (PET tomography) to that in blood (blood samples) (Fig. 1) was analyzed by kinetic modeling of the metabolism of FDG (Fig. 2). Calculations were performed using the simplified Gjedde-Patlak linearization procedure,20–21 which is applicable to liver FDG kinetics using arterial blood input.22 This model representation quantifies the irreversible metabolic processes, and the net metabolic clearance of FDG, K, denotes net transport of FDG from blood into the cell and intracellular conversion to FDG-6-phosphate.23
Visual analysis was performed using both static scan images and parametric images for determining the approximate anatomical location of focal lesions, and the parametric images were used to assign lesions as malignant or nonmalignant using a priori defined criteria based on our previous experience as follows18: a focal area with a lesion/surrounding tissue ratio of K ≥3 was ascribed to malignancy, whereas focal areas with K ratios <3 or with clearly longitudinal configuration compatible with biliary inflammation or stents were ascribed to nonmalignant changes, most probably being local cholangitis. For all patients there was complete agreement between independent assessments by two experienced readers (S. K., O. L. M.).
Enrollment of patients was stopped before the intended 30 patients were included because of slow patient recruitment (approximately 1 of every 3 patients who were offered to participate in the study declined).
The individual tumor-relevant clinical data, PET results, and histological findings of all 24 patients are shown in Table 2. Twenty-three patients underwent transplantation: 16 in Gothenburg, 5 in Stockholm, and 2 in Copenhagen. The median time interval between PET examination and transplantation was 37 days (range 1-307) (Table 2). One patient (patient 11) was taken off the waiting list 96 days after PET due to development of CC.
Histological examination of the 23 explanted livers and biopsy tissue from the patient taken off the waiting list showed PSC cirrhosis in 19 patients and PSC with fibrosis in 5 patients. In addition to patient 11, who was taken off the list because of CC, 2 patients (patients 6 and 9) had CC; 1 patient (patient 7) had high-grade dysplasia in the hilar bile ducts (Table 2).
Comparison of the PET results with the histological results was performed after all 23 patients had been transplanted. The results are shown in Tables 2 and 3. PET was negative in 19 of 20 patients without CC (true negative) and positive in each of the 3 patients with CC (true positive), with mean K ratios >5. In 1 patient with PSC-related cirrhosis and epitheloid cell granulomas in the explanted liver (patient 8), PET was false positive, depicting two small, distinct lesions under the surface of the liver with K ratios of 4 and 8. In patient 7, who had high-grade dysplasia in hilar bile ducts but no CC in the explanted liver, PET gave false negative results.
Table 3. Comparison of Results of Dynamic FDG-PET Prior to Liver Transplantation and Histopathology of Explanted Livers
The sensitivity of PET for diagnosing CC or high-grade hilar dysplasia was 0.75 (95% CI 0.19-0.99) and the specificity was 0.95 (95% CI 0.75-0.99) (Table 3). The negative predictive value of PET for CC or high-grade dysplasia was 0.95 (95% CI 0.75-0.99) and the positive predictive value was 0.75 (95% CI 0.19-0.99) (Table 3). Because pretransplantation CT and ultrasonography results were negative in all patients, the negative predictive value of these modalities was 0.83 (95% CI 0.63-0.95).
Follow-up after transplantation ranged from 10 to 62 months (Table 2). Patient 6, who had CC diagnosed in the explanted liver, and patient 7, who had high-grade dysplasia in the hilar bile ducts, are both alive and well with no evidence of CC nearly 4 years after transplantation. Patient 9, who during transplantation was considered to not have extrahepatic spread of her CC, succumbed to disseminated CC 7 months after transplantation. Patient 11, who was withdrawn from the waiting list because of development of CC during waiting time, died 3 months later from disseminated CC. Patient 14, who had high-grade atypical biliary epithelial cells on preoperative brush cytology, had no evidence of CC in the explanted liver; however, 19 months after transplantation, he developed CC stricturing of the duodenum and died from disseminated CC 28 months after transplantation. Patient 18 died 28 days postoperatively due to sepsis and multiorgan failure. The remaining 18 patients are alive 10 to 62 months after transplantation, with no evidence of CC.
In this study of patients with PSC, dynamic FDG-PET correctly identified 3 of 3 CCs but failed to detect 1 case of high-grade hilar bile duct dysplasia. None of these 4 lesions was seen on US or CT images. In 20 patients without malignancies, there was 1 false positive PET result in a patient with epitheloid granulomas. Our data thus support the hypothesis that PET is superior to conventional imaging techniques for both detection and exclusion of CC in advanced PSC. However, the results must be interpreted cautiously because of the relatively low number of patients in the study—a consequence of PSC being an infrequent disease and the difficulty of obtaining a series of patients with PSC treated via liver transplantation. Nevertheless, the study yielded consistent results.
The use of dynamic PET scanning allows for evaluation of both static scan images of mean radioactivity concentrations 60 to 90 minutes after FDG injection and parametric images of K (i.e., the net metabolic clearance of FDG), being an established measurement of the rate of glucose metabolism.18, 20–22 For conventional static PET images, it is customary to attribute focal areas with lesion/surrounding tissue radioactivity concentration ratios >2 to malignancy.12 We used lesion/surrounding tissue ratios of the K values in the parametric images to discriminate between malignant and nonmalignant changes, ascribing lesions with K ratios >3 to malignancy, based on previous experience.18 In each of the present cases of CC, the lesion/surrounding liver tissue ratio was substantially higher in the parametric images than in the static images (Fig. 3 and Fig.4). In some of the static scan images there were small areas with lesion/surrounding tissue ratios >2 but with K ratios <3. Such lesions were not ascribed to malignancies and were probably due to local cholangitis.
The positive PET result in patient 8 with epitheloid cell granulomas but no malignancy in the liver is a special case. The findings are in agreement with the well-known FDG accumulation in sarcoidosis.24, 25 Both of the two small lesions in this patient were located under the surface of the liver, and both readers considered them as not being typical for CC but categorized them as malignant according to the a priori criteria. Due to the blinded study design, the PET findings were not known to the clinicians, but in clinical practice such findings would probably have lead to laparoscopy with biopsies, whereby the findings would have been classified as nonmalignant. This would change the results shown in Table 3 in such a way that specificity and positive predictive value would both increase to 1.0, whereas sensitivity and negative predictive value would remain the same.
The findings of this study support the use of dynamic FDG PET for sensitive pretransplantation detection of CC in patients with advanced PSC. In this context, it may be mentioned that measurements of radioactivity concentrations in arterial blood samples may be replaced by the time course of the blood radioactivity concentrations in aorta extracted from the dynamic scans,26 making the procedure totally noninvasive. Furthermore, the increasing use of combined PET and CT cameras has the additional advantage of allowing precise anatomical localization of possible malignant lesions in one combined scanning procedure.
Fevery et al.27 recently reported the results of FDG-PET examinations in 10 consecutive pretransplantation PSC patients. Three patients had pretransplantation CC evident on CT or MRI that was later confirmed histologically and were PET-positive (true positive). The remaining 7 patients had negative pretransplantation CT or MRI and thus are comparable with our patients: PET results were true negative in 4 patients, false positive in 2 patients, false negative in 1 patient, and true positive in no patients. Only 1 patient had CC; this patient had false negative PET results, but the time interval between PET and transplantation in this case was 120 days. Fevery et al. found false positive PET results in 2 of 7 patients compared with 1 of 24 patients in the present study. Differences in PET methodology might explain these discrepancies. Our use of dynamic scanning with arterial blood measurements permitted differentiation between lesions with very much increased glucose metabolism ascribed to malignancy (lesion/liver tissue K ratio >3) and lesions with less increased of FDG metabolism ascribed to nonmalignant lesions such as cholangitis. Fevery et al. do not report their PET methodology, but if they used traditional static scanning, false positive findings might have been due to cholangitis. Moreover, it is not evident whether the PET and histology data were evaluated independently.27
The present results indicate that dynamic FDG-PET scanning is superior to CT and US imaging techniques for detecting and excluding CC in advanced PSC. Thus, each of the 3 cases with CC were seen on PET but not on CT. Two of these patients died during the follow-up period, and 1 patient is alive and with no evidence of recurrent CC 46 months after transplantation. The only false positive case (epitheloid cell granulomas) was classified as such according to a priori criteria; however, in a routine clinical setting, the PET findings would have been verified by, for example, laparoscopic examination, and would have been classified as nonmalignant. The only malignant case in the present study missed by PET had high-grade bile duct dysplasia that was eventually found on pretransplantation brush cytology; the patient is alive with no evidence of CC 47 months after transplantation. These observations are in agreement with a recent report by Boberg et al.28 of a high diagnostic accuracy of brush cytology for CC in situ in patients with PSC with dominant bile duct strictures.
In conclusion, pretransplantation dynamic FDG PET of patients with PSC may be a valuable supplement to standard workup programs and may improve management of the patients.
Members of The Swedish Internal Medicine Liver Club: Sven Almer, University Hospital, Linköping; Annika Bergquist, Ulrika Broomé, and Hans Glaumann, Karolinska University Hospital, Huddinge; Einar Björnsson, Rolf Olsson, and Sven Wallerstedt, Sahlgrenska University Hospital, Gothenburg; Åke Danielsson, University Hospital, Umeå; Rolf Hultcrantz, Karolinska University Hospital, Solna; Stefan Lindgren and Hans Verbaan, University Hospital, Malmö; Lars Lö öf, Center for Clinical Research, Västerås; Hanne Prytz, University Hospital, Lund; Hanna Sandberg-Gertzén, University Hospital, Örebro; Per-Gunnar Sangfelt, Uppsala Academic Hospital, Uppsala.